The avoidance of train collisions is vital for human safety in railway transportation. Technical approaches are general train control or collision avoidance systems as well as semi-automated or fully autonomous trains. These systems rely on robust and exact train localization as well as an accurate map of the track network. We present Simultaneous Localization and Mapping relying exclusively on train-side sensors. RailSLAM, implemented as a probabilistic filter, uses measurements from multiple sensors and computes a track map. We rely heavily on sensors that are not affected by the harsh environmental conditions often experienced in this application, in particular a low-cost MEMS Inertial Measurement Unit (IMU). Rail vehicle localization methods based on these sensors require a dedicated map with detailed geometric track features in combination with the topological track connections. If this feature map does not exist apriori, it needs to be created. If it does, it may suffer from incompleteness, insufficient accuracy or outdated information. RailSLAM addresses the creation and maintenance of this special track map by a simultaneous estimation of the probabilistic geometric-topological feature-rich track map and the train state. A first proof of concept implementation of mapping is given based on the use of an Extended Kalman Filter with measurements from Global Navigation Satellite System (GNSS) and an IMU.
展开▼
